Explaining market value of players using PDP

I am going to explain how features "Aggression", "Vision" affects players market predictions. I use fifa-23 dataset. Models that I use are Random Forest Regressor and MLP Regressor.

First model that I use is RandomForestRegressor. I have already used and shortly described in homework 1.

Coefficient of determination: 0.77
Scatter plot: predictions vs truth

Calculating ceteris paribus: 100%|██████████| 62/62 [00:00<00:00, 168.23it/s]

Calculating ceteris paribus: 100%|██████████| 2/2 [00:00<00:00, 55.46it/s]

MLP - multi-layer perceptron is a neural network that consists of at least 3 layers. One layer usually consist of a linear and an activation.

Coefficient of determination: 0.81
Scatter plot: predictions vs truth

Calculating ceteris paribus: 100%|██████████| 2/2 [00:00<00:00, 25.94it/s]

2. Ceteris Paribus explanations are calculated for the Random Forest model, players Messi, Courtois, and features 'Aggression' and 'Vision'. I chose these features at first because they seems to meet task conditions most likely to me.
3. As we can see, increasing 'Aggression' has a positive effect on first player's prediction and negative for second one.
4. Forest's CP claims that increasing 'Aggression' is usually good, but PDP shows that is not for a second player.
5. PDP between MLP and Random Forest are different. MLP's PDP claims that both increasing 'Vision' have linear negative effect on market value prediction and Forest's PDP for 'Vision' and 'Aggression looks more complicated.